Abstract: In this paper,taking Yushu county, Chengduo county and Maduo county in Sanjiangyuan Regions as a case, the Landsat 8 OLI image was used to predict the spatial distribution pattern of topsoil total nitrogen contents. The spectrum reflectance ( R) and its two kinds of transformation forms, including the spectrum reflectance reciprocal (1/ R) and the logarithm of spectrum reflectance reciprocal [lg(1/ R)],selected to relate to soil total nitrogrn measured in laboratory. Firstly, correlation analysis between above three spectral index and the measured topsoil (0-30 cm) total nitrogen was conducted. Secondly, according correlation analysis results, the spectral index with the highest correlation was selected. In the end, the regression models were established using principal component with significant levels of correlated bands. The results show that the spectral reflectance and its two transformation forms from B1-B4, B7 were significantly correlated levels with the measured data, in which the lg(1/ R) was the most obvious. The negative quadratic polynomial model was set up through the first and second principal components of lg(1/ R) of these five bands, in which the R² of calibration model R² was 0.621, RMSE was 2.075, validation samples R² was 0.730, RMSE was 1.493 and RPD was 1.849, suggesting the predicting model having a high precision, good stability. Therefore the OLI images could be used to estimate the spatial distribution pattern of topsoil total nitrogen better.

Key words: soil, total nitrogen, Landsat 8 OLI image, multispectral inversion, Sanjiangyuan area